Premises monitoring using acoustic models of premises

Abstract

According to some embodiments, a control device for a premises security system that is configured to monitor a premises using a plurality of premises devices is provided. The control device is configured to obtain a plurality of acoustic samples for the premises, detect a sound anomaly in at least one of the plurality of acoustic samples, obtain a verification that the sound anomaly is expected, generate an acoustic model for the premises based at least on the plurality of acoustic samples and the verification that the sound anomaly is expected, receive data representing a detected sound during monitoring of the premises, compare the detected sound with the acoustic model for the premises to determine that the detected sound is unexpected, and initiate a premises security system alert based at least on the detected sound being unexpected.

What is claimed is: 
     
       1. A control device for a premises security system that is configured to monitor a premises using a plurality of premises devices, the control device comprising:
 at least one processor; and
 at least one non-transitory computer-readable medium storing a plurality of instructions that, when executed by the at least one processor, cause the at least one processor to: 
 generate a trained acoustic model for the premises by at least: obtaining a plurality of acoustic samples recorded at the premises; 
 detecting a sound anomaly in at least one of the plurality of acoustic samples; obtaining a verification that the sound anomaly is expected; and 
 training an acoustic model based at least on the plurality of acoustic samples and at least by incorporating, into the acoustic model, at least one verified acoustic sample comprising the sound anomaly in response to the verification that the sound anomaly is expected; and 
 implement the trained acoustic model for monitoring the premises, the trained acoustic model comprising at least one unverified acoustic sample and the at least one verified acoustic sample. 
 
 
     
     
       2. The control device of  claim 1 , wherein the plurality of instructions is further configured to cause the at least one processor to:
 receive data representing a detected sound during monitoring of the premises; 
 determine whether the detected sound corresponds to a verified anomaly of the acoustic model; and 
 initiate a premises security system alert based at least on the detected sound not corresponding to the verified anomaly. 
 
     
     
       3. The control device of  claim 1 , wherein:
 each of the plurality of acoustic samples is associated with a respective sample time stamp indicating at least one of a day or a sample period associated with the acoustic sample; and 
 the sound anomaly is associated with a detection time stamp indicating at least one of a day or a time when the sound anomaly was detected. 
 
     
     
       4. The control device of  claim 3 , wherein the plurality of instructions is further configured to cause the at least one processor to compare the detection time stamp and the sample time stamp to determine whether the sound anomaly is expected or unexpected. 
     
     
       5. The control device of  claim 1 , wherein the plurality of instructions is further configured to cause the at least one processor to:
 cause transmission of at least one message to at least one user to prompt the at least one user to classify the sound anomaly as expected or unexpected; 
 receive a response to the at least one message indicating a classification of the sound anomaly; 
 tag the sound anomaly with the classification; and 
 wherein the acoustic model is generated based at least on the classification. 
 
     
     
       6. The control device of  claim 1 , wherein the sound anomaly is associated with at least one of:
 a sound of running water; 
 a sound of door creak; 
 a sound of window creak; 
 a sound of at least one footstep; 
 a sound of a breaking window; or 
 a sound of a pet eating. 
 
     
     
       7. The control device of  claim 1 , wherein at least one of the plurality of acoustic samples is obtained from at least one premises device at the premises, the at least one premises device comprising a sound detector for detecting sound. 
     
     
       8. The control device of  claim 1 , wherein at least one of the plurality of acoustic samples comprises a location stamp indicating a location in the premises where the acoustic sample was generated during a time window; and
 wherein the plurality of instructions is further configured to cause the at least one processor to generate the acoustic model based at least on the location stamp associated with at least one of the plurality of acoustic samples. 
 
     
     
       9. The control device of  claim 1 , wherein the plurality of instructions is further configured to cause the at least one processor to:
 determine that the acoustic model for the premises indicates that a particular sound is expected in the premises; 
 determine that the particular sound has not been detected during monitoring of the premises; and 
 initiate an additional premises security system alert based at least on the particular sound not being detected during monitoring of the premises. 
 
     
     
       10. A method implemented by a control device for a premises security system that is configured to monitor a premises using a plurality of premises devices, the method comprising:
 generating a trained acoustic model for the premises by at least:
 obtaining a plurality of acoustic samples recorded at the premises; 
 detecting a sound anomaly in at least one of the plurality of acoustic samples; 
 obtaining a verification that the sound anomaly is expected; and 
 training an acoustic model based at least on the plurality of acoustic samples and at least by incorporating, into the acoustic model, at least one verified acoustic sample comprising the sound anomaly in response to the verification that the sound anomaly is expected; and 
 implementing the trained acoustic model for monitoring the premises, the trained acoustic model comprising at least one unverified acoustic sample and the at least one verified acoustic sample. 
 
 
     
     
       11. The method of  claim 10 , further comprising:
 receiving data representing a detected sound during monitoring of the premises; 
 determining whether the detected sound corresponds to a verified anomaly of the acoustic model; and 
 initiating a premises security system alert based at least on the detected sound not corresponding to the verified anomaly. 
 
     
     
       12. The method of  claim 10 , wherein:
 each of the plurality of acoustic samples is associated with a respective sample time stamp indicating at least one of a day or a sample period associated with the acoustic sample; and 
 the sound anomaly is associated with a detection time stamp indicating at least one of a day or a time when the sound anomaly was detected. 
 
     
     
       13. The method of  claim 12 , further comprising comparing the detection time stamp and the sample time stamp to determine whether the sound anomaly is expected or unexpected. 
     
     
       14. The method of  claim 10 , further comprising:
 cause transmission of at least one message to at least one user to prompt the at least one user to classify the sound anomaly as expected or unexpected; 
 receive a response to the at least one message indicating a classification of the sound anomaly; 
 tag the sound anomaly with the classification; and 
 wherein the acoustic model is generated based at least on the classification. 
 
     
     
       15. The method of  claim 10 , wherein the sound anomaly is associated with at least one of:
 a sound of running water; 
 a sound of door creak; 
 a sound of window creak; 
 a sound of at least one footstep; 
 a sound of a breaking window; or 
 a sound of a pet eating. 
 
     
     
       16. The method of  claim 10 , wherein at least one of the plurality of acoustic samples is obtained from at least one premises device at the premises, the at least one premises device comprising a sound detector for detecting sound. 
     
     
       17. The method of  claim 10 , wherein at least one of the plurality of acoustic samples comprises a location stamp indicating a location in the premises where the acoustic sample was generated during a time window; and
 the method further comprising generating the acoustic model based at least on the location stamp associated with at least one of the plurality of acoustic samples. 
 
     
     
       18. The method of  claim 10 , further comprising:
 determining that the acoustic model for the premises indicates that a particular sound is expected in the premises; 
 determining that the particular sound has not been detected during monitoring of the premises; and 
 initiating an additional premises security system alert based at least on the particular sound not being detected during monitoring of the premises. 
 
     
     
       19. A control device for a premises security system that is configured to monitor a premises using a plurality of premises devices, the control device comprising:
 at least one processor; and 
 at least one non-transitory computer-readable medium storing a plurality of instructions that, when executed by the at least one processor, cause the at least one processor to:
 generate a trained acoustic model for the premises by at least: 
 
 obtaining a plurality of acoustic samples recorded at the premises; 
 detecting a sound anomaly in at least one of the plurality of acoustic samples; 
 obtaining a verification that the sound anomaly is expected; and
 training an acoustic model for the premises based at least on the plurality of acoustic samples and the verification that the sound anomaly is expected, the training comprising incorporating, into the acoustic model, at least one verified acoustic sample comprising the sound anomaly in response to the verification, the trained acoustic model comprising at least one unverified acoustic sample and the at least one verified acoustic sample; 
 determine that the trained acoustic model for the premises indicates that a particular sound is expected in the premises; 
 determine that the particular sound has not been detected during monitoring of the premises; and 
 initiate a premises security system alert based at least on the particular sound not being detected during monitoring of the premises. 
 
 
     
     
       20. The control device of  claim 19 , wherein at least one of the plurality of acoustic samples is obtained from at least one premises device at the premises, the at least one premises device comprising a sound detector for detecting sound.